Spike–wave discharges (SWDs) in electroencephalography (EEG) are critical biomarkers for diagnosing and monitoring epileptic disorders, particularly absence seizures. Accurate classification of SWDs remains challenging due to their non-stationary, transient, and highly variable time–frequency characteristics. This study proposes a Short-Time Fourier Transform (STFT)-based time–frequency feature extraction framework for automated EEG spike–wave discharge classification. The framework integrates signal preprocessing, STFT-based spectral decomposition, feature engineering, and machine learning-based classification to enhance discriminative performance.

Unlike traditional time-domain or frequency-domain approaches, the proposed method captures localized spectral dynamics, enabling robust representation of transient epileptiform activity. The methodology is conceptually aligned with synchronization and nonlinear EEG behavior studies as highlighted in prior research (Quiroga et al., 2002). Comparative insights from literature demonstrate that time–frequency methods outperform conventional feature extraction techniques in epileptic EEG classification tasks (Tzallas et al., 2009; Martinez-Vargas et al., 2011).

The proposed framework is evaluated conceptually for its ability to differentiate SWDs from normal EEG patterns, emphasizing feature stability, computational efficiency, and clinical interpretability. Results indicate that STFT-based representations significantly enhance classification separability when integrated with machine learning models such as k-NN and ANN. The study further highlights limitations related to window selection sensitivity and computational overhead.

Overall, this research contributes a structured analytical pipeline for EEG spike–wave discharge classification and provides a scalable foundation for real-time neurological diagnostic systems.